Pandas Dataframe，更改“对角线”的值（其中index-value等于column-name）

Question

I have a pandas DataFrame, where the index is a subset of the columns, ie each value in the index is also a column-name and there are additional columns, so something like this:

import pandas as pd
import numpy as np
df = pd.DataFrame(index=['John', 'Mary', 'Steven'],
                  columns=['John','Susan','Steven','Chris','Mary'],
                  data=np.arange(15).reshape(3,5))

I now want to set the "diagonal" items to a fixed value, say 0.0. By "diagonal" I mean those entries, where the index-value matches the column name. I can do it by iterating over the index and set each "diagonal" entry to 0.0 within that loop.

for i in df.index:
    df.loc[i, i] = 0.0

But I wonder if there was a more pythonic way, ie maybe a vectorized of achieving this. I'm thinking that there must be something like df.loc[df.index, df.index] = 0.0 but that does not produce the desired results.

Answer 1

you can use the numpy's fill_diagonal function

import numpy as np 

#rearrange columns according to order of index
df = df.reindex(columns=list(df.index) + list(df.columns.difference(df.index))) 

np.fill_diagonal(df.values,0)

print df
         John  Mary  Steven  Chris  Susan
John       0     4       2      3      1
Mary       5     0       7      8      6
Steven    10    14       0     13     11